effective flooding based on neighbor list exchange...
Post on 25-Aug-2020
1 Views
Preview:
TRANSCRIPT
東海大学大学院2011年度修士論文
Effective Flooding Based on Neighbor List Exchange Over Ad Hoc
Network
指導 石井啓之教授
東海大学大学院工学研究科
情報通信制御システム工学専攻
Turganzhan Kassymov
トゥルガンジャン カスィモフ
i
ABSTRACT
Recently very big earthquake and tsunami occurred in Japan. The
communication infrastructure was out of order at that time, which does
require a new infrastructureless communication environment. Mobile Ad
hoc Network (MANET) must be one of the powerful platform to meet the
requirement. Energy consumption is the critical issue for MANET.
Prolonging a network lifetime without external energy source is one of the
main goals in case of any disasters or emergency situations, because most
of the MANET nodes are driven by battery. So, energy saving MANET is
the key issue. Moreover, at the very beginning of MANET setting up, it is
easily expected that nodes disconnected from the infrastructure have little
information such as IP addresses of other nodes of possible MANET
members. Bearing these points in mind, this thesis proposes two new
broadcasting methods that will reduce power consumption by decreasing
the number of re-broadcastings while keeping almost same packet
penetration rate as in the Simple Flooding (SF) method. These ideas are
based on using the neighboring node information exchange. The
simulation results show the effectiveness of proposals in comparison with
SF.
ii
INDEX
1. INTRODUCTION
1.1 Background
1.2 Existing problems
1.2.1 Energy management
1.2.2 Simple flooding
1.3 Objective of this thesis
1.4 Structure of this thesis
2. RELATED WORK
2.1 Flooding in MANET
2.1.1 Overlay-based approaches
2.1.2 Local-knowledge-based approaches
2.2 MISTRAL
2.3 Conclusion
3. EFNEX (Effective Flooding Based on Neighbor List Exchange
Over Ad Hoc Network)
3.1 Method description
3.1.1 Initialization Phase
3.1.2 Transfer Phase
3.1.3 Packet Structure
3.1.4 Explanation by an example
3.2 Evaluation
3.3 Conclusion
iii
4. EFNEX-R (Effective Flooding Based on Neighbor List
Exchange Over Ad Hoc Network - Recreated)
4.1 Method description
4.1.1 Initialization Phase
4.1.2 Transfer Phase
4.1.3 Packet Structure
4.2 Evaluation
4.3 Conclusion
5. CONCLUSION
6. ACHIEVEMENTS
7. REFERENCES
1
1. INTRODUCTION
1.1 Background
Mobile Ad hoc Network (MANET) is a wireless infrastructure-less
self-configured mobile network that is used for various purposes in virtue
of easy deployment and decentralized network administration. Actually, the
MANET is actively used by military, researchers and students to create a
local network. This network carries out not only information exchange
between local nodes, but also allows connecting to the Internet. And
because the each node can move randomly, a network topology changes at
will.
Because of boneless structure, user can move around the network and get
the same personalized service without relation to changing location,
switching between devices or sessions.
MANET is a kind of special wireless network mode. A MANET is a
collection of two or more devices equipped with wireless communications
and networking capability. Such devices can communicate with another
device that is immediately within their radio range or one that is outside
their radio range not relying on access point. A MANET is
self-organizing, self-disciplining and self-adaptive. The main
characteristics of MANETs are as follows[4]:
(1) Dynamic topology
Because nodes in the network can move arbitrarily, the topology of the
network also changes. The bandwidth of the link is unstrained, and the
capacity of the network is also tremendously variable. Because of the
dynamic topology, the output of each relay node will vary with the time,
and then the link capacity will change with the link change. At the same
time, complete-collision and interference make the actual bandwidth of ad
hoc networks smaller than their bandwidth in theory.
(2) Power limitation in mobile devices
Because of the mobility characteristics of the network, devices use batteries
2
as their power supply. As a result, advanced power conservation techniques
are very necessary in designing a system.
(3) Security
The mobile network is more easily attacked than the fixed network.
Overcoming the weakness in security and the new safety trouble in wireless
networks are on demand.
According to [1] , MANETs can be used to create following:
Community network
Enterprise network
Home network
Emergency response network
Vehicle network
Sensor network
In case of using MANET, the network can be established in hours as
alternative of weeks that is needed to make robust wired network.
Besides, in case of any emergencies, when public communication services,
such as wired and cellular networks, are out of order, MANET can be
quickly deployed for connecting rescue teams and victims. Ad-hoc network
usually consist of a number of mobile nodes. It can be laptops or cellular
phones that are in broadcasting area of each other. Advantages of MANET
are that there is no need of network infrastructure and base stations and that
each terminal acts as receiver, transmitter and re-transmitter. Development
of MANET is one of important parts of self-defense organization in Japan
because of frequent natural disasters.
3
1.2 Existing problems
1.2.1 Energy management
Besides of advantages mentioned in 1.1, MANET often faces with some
critical issues of a network lifetime prolonging and a network penetration
rate. The reason of lifetime problems is that often a public power
infrastructure gets down too, and the ad-hoc terminals don’t have external
energy source, only an existing battery charge. The problems with the
penetration rate are connected with a node location and natural or artificial
barriers neighboring the node.
In accordance with [4], the main reasons for energy management in
MANETs are as follows:
(1) Limited energy reserve
The MANETs have limited energy reserve. The improvement in battery
technologies is very slow as compared to the advances in the field of
mobile computing and communication.
(2) Difficulties in replacing the batteries
In situations like battlefields, natural disasters such as earthquakes, and so
forth, it is very difficult to replace and recharge the batteries. Thus, in such
situations, the conservation of energy is very important.
(3) Lack of central coordination
Because an ad hoc network is distributed network and there is no central
coordinator, some of the nodes in the multi hop routing should act as relay
node. If there is heavy relay traffic, this leads to more power consumption
at the respective relay node.
(4) Constraints on the battery source
The weight of the nodes may increase with the weight of the battery at that
node. If the weight of the battery is decreased, that in turn will lead to less
power of the battery and thus decrease the life span of the battery. Thus,
energy management techniques must deal with this issue; in addition to
4
reducing the size of the battery, they must utilize the energy resources in
the best possible way.
(5) Selection of optimal transmission power
The increase in the transmission power increases the consumption of the
battery charge. Because the transmission power decides the reachability of
the nodes, an optimal transmission power decreases the interference
between nodes, and that in turn increases the number of simultaneous
transmission.
Figure 1.2.1 provides an overview of the network energy management
schemes.
Figure 1.2.1 Classification of energy management schemes
5
1.2.2 Simple flooding
Mobile Ad Hoc Networks (MANETs) are the appropriate candidates to be
used in case of emergency situation because the MANET does not need any
infrastructure which may suffer from unavailability due to disaster. Just
after the disaster, one of the most necessary functionalities of MANET is
“Broadcasting” over all over the network to let disaster victims know
where to evacuate, where the medical doctors are, where the shelters are,
how their families are.
If the network infrastructure becomes out of order just after the emergency
situation happens, the MANET is a very useful means to keep
communication. But at the very beginning of the MANET setting-up, it is
easily anticipated that each node in the MANET has such little knowledge
about other node as IP address. The most powerful means at that phase is
“Broadcasting” which requires no specific destination addresses. The most
popular broadcasting method is “Simple Flooding”. Broadcasting is
organized as follows:
(1) An initiator node broadcasts the packet with a data.
(2) Each node that received this packet re-broadcasts it.
(3) If the packet had been received before and re-broadcasted, node doesn’t
carry out the re-broadcasting.
Unfortunately, flooding has been shown to be susceptible to contention
even in reasonably dense networks. Indeed, flooding leads to a large
amount of redundant messages that consume scarce resources such as
bandwidth and power and cause contention, collisions and thus additional
packet loss. Every node receives the message from every neighbor within
transmission range, except when messages are lost due to contention and
collisions . This fact will cause excessive battery consumption leading the
network to shorter lifetime. This problem is known as the broadcast storm
problem. Because flooding is important in MANET applications, there is a
clear need for storm-resistant flooding protocols that operate efficiently.
However, reducing the number of redundant broadcasts leads to a lower
degree of reliability. Hence, the challenge we face is to strike a balance
6
between message overhead (i.e., the level of redundancy) and reliability .
When it is used, redundant re-broadcasts occur, increasing the chances of
collisions and buffer overflows, and resulting in a considerable degradation
in delivery quality. In case a packet is sent by an initiator node, all the
nodes will re-broadcast once. As shown in Figure 1.2.2, if there are 30
nodes in the ad hoc network, 30 packets are generated which are
completely identical. Hence, many packets are generated continuously,
so-called flood is to be occurred and in some cases, collision and loss will
be caused. And redundancy will cause excessive use of battery.
Figure 1.2.2 «Flood» in Simple Flooding
7
1.3 Objectives of this thesis
Making network lifetime without external energy source as long as possible
is one of the main goals in case of any disasters or emergency situations.
This research aims at proposing a new broadcasting method that allows
reducing power consumption by decreasing a number of redundant
re-broadcastings with keeping almost same packet penetration rate as in the
SF method.
8
1.4 Structure of this thesis
This thesis consists of 5 chapters. After Introduction, in chapter 2, we show
one example of related works that was made to solve existing issues in
MANET. This work is “MISTRAL: Efficient Flooding in Mobile Adhoc
Networks” by Stefan Pleisch, Mahesh Balakrishnan, Ken Birman, and
Robbert van Renesse.
In Chapter 3, this thesis proposes a new method of effective flooding based
on neighbor list exchange, "EFNEX". It is based on exchange of the
information about neighboring nodes. In this case, previous broadcaster
uses a selection method of next broadcaster based on the number of
neighbors. Through simulation, this method allows decreasing of redundant
broadcastings in 2 times, but with insignificant reachability degradation in
3-4% from SF.
In Chapter 4, an additional algorithm of effective flooding, "EFNEX-R" is
proposed, that enables cutting of the redundant broadcastings in 7 times
approximately with 12% of the reachability rate degradation in comparison
with Simple flooding and EFNEX. In this proposed algorithm, the previous
broadcaster selects next broadcaster among its neighboring nodes based on
information about a number and names of their neighbors.
The Chapter 5 closes this thesis. In this part of work we bring advantages
and disadvantages of proposed methods in comparison with other existing
methods. Conclusions and supposed outputs will be determined as well.
9
2. RELATED WORK
MANET is a very popular field of research. And there are many works and
inventions made by wireless communication scientific community. Among
the large amount of papers, the author is interested in “Efficient Flooding in
Mobile Adhoc Networks” by Stefan Pleisch, Mahesh Balakrishnan, Ken
Birman, Robbert van Renesse. In their paper, they propose a novel
approach to flooding, which relies on proactive compensation packets
periodically broadcast by every node. The compensation packets are
constructed from dropped data packets, based on techniques borrowed from
forward error correction. Since this approach does not rely on proactive
neighbor discovery and network overlays it is resilient to mobility.
This thesis clearly explains this algorithm to make understanding in current
MANET trends.
10
2.1 Flooding in MANET
In any flooding mechanism, one must balance reliability against message
overhead. On the one hand, increasing reliability generally involves
sending a greater number of redundant messages and thus incurs a higher
message overhead. In this worst case, the system risks provoking broadcast
storms[5]. Yet redundant messages are needed to reach all nodes and to
recover from packet loss, hence reducing the overhead will generally
decrease reliability.
The broadcast storm problem is so common in flooding algorithms that it
has engendered a whole area of research. Storm-sensitive flooding
approaches can be broadly classified into two classes:
local-knowledge-based and overlay-based. Local-knowledge-based
approaches decide on whether to rebroadcast or drop a flooded message
solely on the basis of local information. Most commonly, they use
information from received broadcasts to adaptively determine the
forwarding policy. Such algorithms are a natural fit for MANETs, as they
do not need to maintain any kind of complex node-to-node state that might
need to be adapted in the event of mobility or other topology changes. In
contrast, overlay-based approaches structure the node field according to
some (local) topology, and then use topological information to efficiently
implement flooding and reliability. The problem here is that if nodes have
low quality connections to neighbors and/or are in motion, the overlay
structure must be adapted. As a consequence, a high rate of management
messages may be required, and if a flooded message is propagated while
the overlay is out of date, that message may experience a high loss rate. In
the worst case, the system might end up in a state of churn, constantly
adapting the overlay but never managing to achieve the high quality of
flooding that the overlay is intended to support.
By briefly overviewing existing work, they are assigned to the
corresponding class. For reasons of brevity, this review is deliberately
partial; focusing on results that inspired our work here, or that have been
widely cited in the literature.
11
2.1.1 Overlay-based approaches
As just indicated, the term overlay is used very broadly. An overlay-based
approaches is an algorithm that superimposes a routing structure onto the
MANET in support of flooding and rebroadcast. Depending on the position
of a node in this overlay, it decides to either rebroadcast a flooded packet,
or to only process and then drop it. While overlays provide a convenient
mechanism to reduce the message overhead of flooding and to increase
reliability, they suffer from the need to reconfigure the overlay when
connectivity changes or if the nodes are mobile. Restructuring adds
overhead but also increases the likelihood that messages will be lost, and
thus may decrease coverage of the flooding protocol.
Ni et al. propose to structure the nodes into clusters. Their solution
rebroadcasts a packet in a manner that depends on the node’s position in
the cluster: only cluster head and gateway nodes rebroadcast.
The goal is to provide low-latency flooding. This is in part achieved by
minimizing the collisions and interference. Gandi et al. show that an
optimal solution to this problem is NP complete, instead, they propose an
approximation algorithm. They construct a multicast tree and compute a
rebroadcasting schedule such that the expected rate of collisions will be
low.
Other approaches are based on the approximation of (minimal) connected
dominating sets (MCDS). Informally, a dominating set (DS) contains a
subset of all nodes such that every node not in the DS is adjacent to one in
the DS. Thus, a DS creates a virtual backbone that can be used to
efficiently flood messages. It has been shown that the creation of an MCDS
is NP-complete. Thus, most approaches attempt to find a sufficiently good
approximation to a MCDS.
A number of approaches rely on two-hop neighbor information to select
nodes that rebroadcast the message. These approaches require that hello
messages containing neighbor information are exchanged between the
nodes.
12
For instance, in the Double-Covered Broadcast (DCB), node n collects
information about the two-hop neighbor set. Among its one-hop neighbors
it then picks nodes that rebroadcast the message (called forward node) such
that (1) the rebroadcast by the forward node covers the two-hop neighbors,
and (2) the one-hop neighbors that are no forward nodes are within range of
at least two rebroadcasts by forward nodes. The reception of the message
by the forward node is implicitly acknowledged when n overhears the
rebroadcast.
The scalable broadcast algorithm (SBA) also uses two-hop neighbor
knowledge, but employs a different approach to select the forward nodes.
With node mobility, the two-hop neighbor sets need to be updated
frequently. Otherwise, the neighbor sets become outdated and reliability
drops.
13
2.1.2 Local-knowledge-based approaches
Local-knowledge-based approaches generally decide on a per-node basis
whether to rebroadcast a particular flooded message. In the simplest case,
each node flips a coin and rebroadcasts messages with a certain
probability p. This approach purely is probabilistic flooding (PPF).
There are a number of variants on this basic idea. For example, one set of
algorithms base the rebroadcast decision either on the number of already
overheard rebroadcasts, or on the distance or location of the overheard
rebroadcast’s sender. The idea underlying these schemes is that the
additional coverage gained by rebroadcasting decreases with the number
of overheard rebroadcasts and decreasing distance to neighboring
rebroadcasting nodes. However, it takes time to collect these statistics,
delaying the rebroadcast decision, hence a potentially high latency is
introduced to every flooded message. In his work, Tseng et al. extend
earlier approaches to allow nodes to dynamically adapt threshold values
such as the rebroadcast counter.
Zhang and Agrawal propose an approach that is a combination of the
counter-based and probabilistic methods. Instead of using a static
rebroadcast probability p, they adjust p according to the information
collected by the counters. While this makes p adaptable, it becomes
dependent upon other fixed parameters that need to be carefully
selected (e.g., timeouts).
Dynamic Gossip relies on local density awareness to adjust the
rebroadcast probability p of the one-hop neighbors. Its correctness and
suitability relies on the assumption that the nodes are uniformly
distributed. Density information is collected using a relay-ping method.
In his work, Kowalski and Pelc propose a broadcasting algorithm with
optimal lower bounds in their model. They consider only stationary nodes
and adjust the broadcast probability accordingly.
Haas et al. study what they term a phase transition phenomena. This work
14
shows that purely probabilistic flooding in an ad hoc network has a
bimodal delivery distribution. Their simulations reveal that either almost
every node receives the message, or virtually none. To reduce the
likelihood of the latter case, they explore a variety of approaches, such as
adapting the rebroadcast probability to the density or the distance to the
flooding source. Sasson et al. theoretically explore the same phenomena
based on percolation theory and conclude that there exists a threshold ¯p
< 1 such that for any p > ¯p the node coverage is close to 1, while for p <
¯p the coverage is very low. Hence, increasing p much beyond ¯p is not
very useful.
Any approach that bases rebroadcast decision on observation of
neighbors and on overheard broadcasts is at risk of using stale
information if nodes might move before the information is used.
MANETs, of course, can have a high degree of mobility, hence neither of
these approaches is ideal.
Mistral’s compensation mechanisms is orthogonal to these approaches.
Indeed, were we building a production deployment of flooding in a
real-world setting, we would be inclined to combine Mistral with one of
these others (as should be clear, the ideal choice of underlying
mechanism depends upon the anticipated density of nodes and level of
mobility; no single solution stands out as uniformly superior to the
others). By using such a hybrid scheme, we could parameterize the
underlying solution to keep overheads low, accepting a modest risk that
flooded packets would fail to reach some nodes. Compensation packets
could then be used to overcome this low level of residual losses.
15
2.2 MISTRAL
Traditional flooding suffers from the problem of redundant message
reception, once per neighbor. Even in a reasonably connected network, the
same message is received multiple times by every node, which is
inefficient, wastes valuable resources, and can create contention in the
transmission medium.
Selective rebroadcasting of flooded messages is a way to limit the number
of redundant transmissions. Instead of simply rebroadcasting the message a
node evaluates a local function F and then uses the outcome of this
computation to decide whether to forward the message. In its simplest form,
this function returns its result based on some static probability
(corresponding to PPF). More complex functions take into account
additional topological (e.g., the number of neighbors) or statistical
information (e.g., the number of overheard rebroadcasts). The downside of
selective flooding is that a flooding may no longer reach all intended nodes.
In particular, if a node has only few neighbors, none of these neighbors
may rebroadcast the message. Selective flooding thus balances message
overhead against reliability.
Mistral finds some middle ground by introducing a new mechanism that
allows us to fine-tune the balance between message overhead and
reliability. The key idea is to extend selective flooding approaches by
compensating for messages that are not rebroadcast. This compensation is
based on a technique borrowed from forward error correction (FEC). Every
incoming data packet (dp) is either rebroadcast or added to a compensation
packet (cp). The compensation packet is broadcast at regular intervals and
allows the receivers to recover one missing data packet.
16
2.3 Conclusion
In this chapter, existing effective flooding methods are reviewed briefly,
mainly focusing on MISTRAL, which is a good basis for discussion in this
thesis.
17
3. EFNEX (Effective Flooding Based on Neighbor List Exchange
Over Ad Hoc Network)
3.1 Method Description
As described so far, MANET is needed in case of emergency and most
convenient way of information distribution is broadcasting, that is flooding,
just after the emergency happens.
However, existing Simple Flooding (SF) causes redundant message
transmission that results in performance degradation and waist of finite
battery resources. To solve these problems, this thesis proposes a new
effective flooding algorithm, named “Effective Flooding based on neighbor
exchange (EFNEX)” .
In this method, a previous sender selects the node that has biggest number
of neighboring nodes as next broadcaster. The node will never broadcast
the message twice. For these purposes, the packet consists of a data that
need to send, a packet ID and names of previous broadcasters and name of
next broadcaster.
18
3.1.1 Initialization Phase
Initialization phase of this procedure is as follows:
1. Each node broadcast a hello message with its name.
2. Each node that received hello message records the name of
broadcasting node and make own neighbor list (NL) to count the number of
neighboring nodes.
3.
[Note] NL contains names of all nodes within a radio range that sent hello/
response massages to this node and the number of these nodes’ neighbors.
19
3.1.2 Working Phase
1. The node checks its NL to find and selects a neighboring node that has
biggest number of neighbors in its NL.
2. The node appoints selected node as next broadcaster by including its
name in a sending packet.
3. The node broadcasts a packet.
4. Each node that received the packet checks the name of next
broadcaster. And if name is equal to its own name, the node repeats
procedures 1-3 above. And reminds that it is obligatory to rebroadcast
all messages coming from that node in case of any additional
instructions absence.
20
3.1.3 Packet Structure
The sending packet consists of following parts:
(1) Data: data is information that needs to be sent.
(2) Names of previous broadcasters: there is a list of all previous
broadcasting nodes. Each broadcaster adds its name to the list names of
previous broadcasters. It enables to determine and exclude the neighboring
nodes, which have broadcasted this packet before, and then make the
selection of next broadcaster.
(3) Packet ID: packet name that needs to determine sequence of receiving
and re-broadcasting packets.
(4) Name of the next broadcaster: Neighboring node’s name that selected
as next broadcaster.
Figure 3.1.1 shows a possible packet structure.
Figure 3.1.1 Packet structure
After transferring pass creation there is no need to include (2) and (4). And
the packet is sent without additional optional information, it enables to bit
decrease a packet size.
Packet Data Names of
previous
receivers
(NLs)
Next
broadcaster
Packet ID
21
3.1.4 Explanation by an example
Consider that there is a node Q. The node Q has neighboring nodes K, S, P,
G expressed in Q’s NL. Also Q node knows names of neighbors of K, S, P,
G.
Node K has S, P, A neighboring nodes.
Node S – K, P, G, A, L, W.
Node P – G, N, T, D, Y.
Node G – S, G.
To select next broadcaster of a packet, Q compares its NL with its
neighbors NL to determine the number of neighboring nodes. Table 3.1.1
expresses NL:
Table 3.1.1 Neighbor list of node
Number of
neighbors
Q K S P G
3 K S P A
6 S K P G A L W
5 P G N T D Y
2 G S P
As you can see, node S has the biggest number of neighboring nodes. Node
S is selected by node Q as a next broadcaster. After selection, node Q
broadcasts the packet with a name of the next broadcaster and list of
previous broadcasters. This method of selection of next broadcaster is used
further.
22
3.2 Evaluation
To evaluate our proposal, a trial simulation on Excel was executed.
The simulation conditions are:
Number of nodes: 100 randomly located immobile nodes
Network area: 1,000m*1,000m
Radio area: 200m radius
After 5 times trials, we have got following results.
Table 3.2.1 shows evaluation results of EFNNEX with comparison with
SF.
Table 3.2.1 Evaluation Results
Type of broadcasting SF EFNEX
Initialization phase 0 200
(2 hellos/node)
Aver. penetration rate 100% 94.7%
Aver. number of broadcastings
per packet
100 51
This result shows that our proposal can reduce the number of
rebroadcasting down to a half of that of Simple Flooding for every packet
sent by a source. Although there is some overhead before starting the
procedure, the reduction effect becomes very large if the continuously sent
packets become large. Table 3.2.2 expresses growth of number of
broadcastings per packet, which shows effectiveness of our proposal.
Table 3.2.2 Growth of number of broadcastings per packet
# of packets 1 10 100 1,000
SF 100 1,000 10,000 100,000
EFNEX 251 710 5,300 51,200
The defect is degradation of packet penetration rate.
23
3.3 CONCLUSION
A new effective broadcasting method "EFNEX" is proposed and evaluated
through a simulation that results in very high effectiveness (almost 50%
decreasing of broadcastings) with relatively low degradation in penetration
rate.
The simulation method considers no physical collision, no moving of
terminals so that computer network simulation may be needed for further
study item.
24
4. EFNEX-R (Effective Flooding Based on Neighbor List
Exchange Over Ad Hoc Network - Recreated)
4.1 Method description
To solve the Simple Flooding problem, we propose a new algorithm,
“Effective Flooding based on neighbor list exchange (EFNEX)" in the
previous chapter that allows reducing the power consumption of network
by cutting the number of the rebroadcastings.
Here, this thesis proposes the refined EFNEX named “Effective Flooding
based on neighbor list exchange - Recreated (EFNEX-R)". In EFNEX,
only the number of NL is used for decision of next broadcaster. Here we
use the names of neighbor nodes.
It is achieved when each node generates a list of neighboring nodes (NL)
and exchanges its NL with all neighboring nodes in initialization phase of
network. After NL exchanging each node knows not only names of
neighboring nodes, but also names of its all neighboring nodes in current
network position. When node wants to broadcast a packet of data to all
existing network, the packet is re-broadcasted only by node that has biggest
number of different neighbors from its neighbors. Thus, a packet is
re-broadcasted only by one node selected by previous broadcaster. This
method keeps out from redundant messages transmission that leads to
network overloading and huge power consumption.
First of all let’s define what “different neighbor node” means:
Let’s imagine there are a number of nodes in local area. And there are node
A and node B, which are in broadcasting radius of each other. We call them
as neighboring nodes or simply neighbors. Each node has a number of
neighboring nodes. And besides each other, A and B have some nodes in
their broadcasting area. And these nodes which are in broadcasting area of
A and B both, we define as similar nodes. Otherwise, each of similar nodes
is neighboring to A and B together. Also A has a number of nodes, which
are only in its broadcasting area in comparison with neighbors of node B.
25
Node B also has such kind of neighboring nodes. We call them as different
neighboring node. Thus, each node has the similar and different
neighboring nodes with each neighboring node.
26
4.1.1 Initialization Phase
(1-1) Each node broadcast a hello message with its name.
(1-2) Each node that received hello message records the name of
broadcasting node and made own neighbor list (NL).
(1-3) Each node broadcasts its NL.
For clear understanding let’s consider next example:
One node wants to create network, but there are not any nodes around it.
The node sends Hello message with its names and waits for response from
other nodes. After receiving responses, node generates own NL using an
information from respond messages. In this case, NL sets up ZERO,
because nobody responded.
After a while, another node wants to create or enter the network. As before,
this node broadcasts Hello message with its name and waits for response.
After receiving hello message from another node, first node broadcasts
response message. Response message includes names of requesting and
responding nodes and responding node’s NL. After receiving the response
message, second or requesting node creates its NL and broadcast it. The
first of responding node updates its NL by adding the information from
second node’s NL.
Initialization phase enables nodes to create its NL that contains names of
all nodes within a radio range that sent hello messages to this node and
names of these nodes’ neighbors. Information in NL helps node to
determine next broadcaster.
27
4.1.2 Working Phase
(2-1) before a packet is broadcasted by a initiating node, the node compares
the its own NL with all the neighbors’ NLs and selects a node as a next
broadcaster that has biggest number of different neighbors from its own
neighbors
(2-2) The initiating node creates the “list of previous receivers” by adding
/merging its neighbors’ names.
(2-3) The initiating node broadcasts a packet, which contains data, a packet
ID, a name of next broadcaster selected in (2-1), list of previous receivers.
(2-3) The next broadcaster compares its neighbors and list of previous
receivers to determine the neighboring nodes, which have received this
packet before. These nodes can’t be considered as next broadcaster;
(2-4) the procedures (2-1) through (2-3) are repeated regarding an initiator
as previous broadcaster.
[note] Within the procedure, nodes selected at once are never selected.
Node’s NL includes list of neighbors except this node.
Figure 4.1.1 Next broadcaster selection method (EFNEX-R)
28
4.1.3 Explanation by an example
Please consider that there is a node Q. The node Q has neighboring nodes
K, S, P, G. that expressed in Q’s NL. Also Q node knows names of
neighbors of K, S, P, G.
Node K has S, P, A neighboring nodes.
Node S – K, P, G, A, L, W.
Node P – G, N, T, D, Y, M.
Node G – S, G.
To select next broadcaster of a packet, Q compares its NL with its
neighbors NL to determine the number of different nodes. Table 4.1.1
expresses NL:
Table 4.1.1 Neighbor list of node
Different
nodes with Q
Similar
nodes with Q
Q K S P G
1 2 K S P A
3 3 S K P G A L W
5 1 P G N T D Y M
0 2 G S P
As you can see, node P has the biggest number of different nodes. Node P
is selected by node Q as next broadcaster. After selection, node Q
broadcasts the packet with a name of the next broadcaster and list of
previous receivers. This method of selection of next broadcaster is used
further.
29
4.1.4 Packet Structure
The sending packet consists of following parts:
[1] Data: data is information that needs to be sent.
[2] Names of previous receivers: there is a list of all previous broadcasters’
neighboring nodes. Each broadcaster merges its neighbors’ names with the
names of previous broadcasters to generate this list. It enables to determine
and exclude the neighboring nodes, which have received this packet before,
and then make the selection of next broadcaster.
[3] Packet ID: packet name that needs to determine sequence of receiving
and re-broadcasting packets.
[4] Name of the next broadcaster: Neighboring node’s name that selected
as next broadcaster.
After transferring pass creation there is no need to include [2] and [4]. And
the packet is sent without additional optional information, it enables to bit
decrease a packet size.
30
4.2 Evaluation
In order to evaluate our proposal, a pilot simulation on Excel was executed.
The simulation conditions are:
Number of nodes: 100 randomly located immobile nodes
Network area: 1,000m*1,000m
Radio area: 200m radius
After 5 times trials, we have got following results.
Table 4.2.1 shows evaluation results of EFNEX-R with comparison with
SF and EFNEX . Table 4.2.2 expresses growth of number of broadcastings
per packet.
Table 4.2.1 Evaluation Results
Type of broadcasting SF EFNEX EFNEX-R
Initialization phase 0 200
(2 hellos/node)
200
(2 hellos/node)
Aver. penetration rate 100% 94% 88.2%
Aver. number of
broadcastings per packet
100 51 15,2
Table 4.2.2 Growth of number of broadcastings per packet
# of packets 1 10 100 1,000
SF 100 1,000 10,000 100,000
EFNEX 251 710 5,300 51,200
EFNEX-R 215 350 1,720 15,400
31
4.3 Conclusion
From the results, it is clear that the proposed method allows cutting the
number of re-broadcastings down to cover network. It means that fewer
nodes involved in sending message, and other nodes can save its power.
This way provides longer network lifetime than Simple Flooding and
EFNEX
The simulation method considers no physical collision, no moving of
terminals so that computer network simulation may be needed for further
study item.
32
5. CONCLUSION
This thesis tried to clarify appropriate network configuration in case of
emergency when existing infrastructure does not work well. The most
powerful candidate to that situation is Mobile Ad hoc Network (MANET).
This thesis focuses on the case where the MANET is introduced just after
the crisis happens, where sufficient information like IP addresses to
construct a unicast-oriented network is not well known to nodes in the
MANET. In this situation broadcasting is the only tool to distribute
information all over the network without use of IP addresses.
But existing broadcasting, namely Simple Flooding (SF) may produces
many redundant duplicated packets that cause the congestion, loss, delay
and waist of battery resources of each node.
So, this thesis tries to propose new methods of efficient flooding that
reduces redundant reproduction of packets and so reduces waist of battery.
This thesis consists of 5 chapters. After Introduction of chapter 1, in
chapter 2, we showed one example of related works that was made to solve
existing issues in MANET. This work is “MISTRAL: Efficient Flooding in
Mobile Adhoc Networks” , which is a good basis for discussion in this
thesis.
In Chapter 3, this thesis proposed a new method of effective flooding based
on neighbor list exchange, "EFNEX", which is based on exchange of the
information about neighboring nodes. In this case, previous broadcaster
uses a selection method of next broadcaster based on the number of
neighbors. Through simulation, this method allows decreasing of redundant
broadcastings in 2 times, but with insignificant reachability degradation in
3-4% from SF.
In Chapter 4, an additional algorithm of effective flooding, "EFNEX-R"
was proposed, that enables cutting of the redundant broadcastings in 7
times approximately with 12% of the reachability rate degradation in
33
comparison with Simple flooding and EFNEX. In this proposed algorithm,
the previous broadcaster selects next broadcaster among its neighboring
nodes based on information about a number and names of their neighbors.
The Chapter 5 concludes this thesis. In this part of work we bring
advantages and disadvantages of proposed methods in comparison with
other existing methods.
34
6. ACHIEVEMENTS
[1] K. Utsu, H. Sano, T. Kassymov, H. Nisikawa, H. Ishii, “Proposal on
Battery-aware Counter-based Flooding over Ad Hoc Networks”, the
2011 International Conference on Parallel and Distributed Processing
Techniques and Applications (PDPTA’11), Jul. 2011
[2] T. Kassymov, K. Utsu, H. Sano, N. Morita, H. Ishii, “Effective Flooding
based on neighbor list exchange over Ad Hoc Networks”, IEICE
Society Conference 2011, Sep. 2011.
35
7. REFERENCES
[1] Utsu, Keisuke & Ishii, Hiroshi (2010) Load-aware Flooding over Ad
Hoc Networks enabling high message reachability and traffic. ICMU
2010 Proceedings of the 5th
International Conference on Mobile
Computing and Ubiquitous Networking
[2] Utsu, Keisuke & Nishikawa, Hiroaki & Ishii, Hiroshi (2010)
Load-aware Flooding over Ad Hoc Networks achieving a reduction in
traffic and power consumption. Las Vegas, USA: PDPTA’10
Proceedings of the 2010 International Conference on Parallel and
distributed Processing Techniques and Applications
[3] Murthy, C. Siva Ram & Manoj, B. S. (2004) Ad Hoc Wireless Networks
- Architectures and Protocols. USA: Prentice Hall
[4] Sarkar, Subir Kumar & Basavaraju, T.G. & Puttamadappa, C. (2008) Ad
Hoc mobile wireless networks: principles, protocols and applications.
Boca Raton, FL: Auerbach Publications
[5] Pleisch, Stefan & Balakrishnan, Mahesh & Birman, Ken & Van Renesse,
Robbert (2006) MISTRAL: Efficient Flooding in Mobile Adhoc
Networks. New York, NY, USA: MobiHoc '06 Proceedings of the 7th
ACM international symposium on Mobile ad hoc networking and
computing
[6] Williams, Brad & Camp, Tracy (2002) Comparison of Broadcasting
Techniques for Mobile Ad Hoc Networks, Proceedings of the 3rd ACM
International Symposium on Mobile Ad Hoc Networking and
Computing, pp. 194-205,
top related